autointent.custom_types#
Types used throughout AutoIntent library.
Attributes#
Float value between 0 and 1, inclusive. |
|
Type alias for label representation |
|
Any label that autointent modules can handle. |
|
List of labels that AutoIntent modules can handle. |
|
List of labels without OOS-samples that AutoIntent modules can handle. |
|
List of labels with OOS-samples that AutoIntent modules can handle. |
|
List label for multi-label classification problems. |
|
List label for multi-label classification problems with OOS samples. |
|
Hyperparameter tuning strategies: |
|
Some presets that our library supports from lightest to heaviest: |
|
Behavior when meet a data-incompatible module in search space: |
|
Integer label for single-label classification problems. |
|
Integer label for single-label classification problems with OOS samples. |
|
Validation scheme used in hyperparameter tuning: |
|
Represents weight calculation methods |
Classes#
Package Contents#
- autointent.custom_types.FloatFromZeroToOne#
Float value between 0 and 1, inclusive.
- type autointent.custom_types.LabelType = SimpleLabel | MultiLabel#
Type alias for label representation
int: For single-label classification.
list[int]: For multi-label classification.
- autointent.custom_types.LabelWithOOS#
Any label that autointent modules can handle.
- autointent.custom_types.ListOfGenericLabels#
List of labels that AutoIntent modules can handle.
- autointent.custom_types.ListOfLabels#
List of labels without OOS-samples that AutoIntent modules can handle.
- autointent.custom_types.ListOfLabelsWithOOS#
List of labels with OOS-samples that AutoIntent modules can handle.
- autointent.custom_types.MultiLabel#
List label for multi-label classification problems.
- autointent.custom_types.MultiLabelWithOOS#
List label for multi-label classification problems with OOS samples.
- autointent.custom_types.SamplerType#
Hyperparameter tuning strategies:
random:
optuna.samplers.RandomSampler
- autointent.custom_types.SearchSpacePreset#
Some presets that our library supports from lightest to heaviest:
light_extra: just like light but tuned with
optuna.samplers.RandomSampler
light: only light-weight modules tuned with
optuna.samplers.TPESampler
light_moderate: just like light but tuned with
optuna.samplers.BruteForceSampler
heavy_moderate: just like heavy but tuned with
optuna.samplers.RandomSampler
light: light-weight modules accompanied with heavy-weight ones tuned with
optuna.samplers.TPESampler
heavy_extra: just like light but tuned with
optuna.samplers.BruteForceSampler
- autointent.custom_types.SearchSpaceValidationMode#
Behavior when meet a data-incompatible module in search space:
raise: raise an error
warning: warn user
filter: drop incompatible modules from search space
- autointent.custom_types.SimpleLabel#
Integer label for single-label classification problems.
- autointent.custom_types.SimpleLabelWithOOS#
Integer label for single-label classification problems with OOS samples.
- autointent.custom_types.ValidationScheme#
Validation scheme used in hyperparameter tuning:
ho: hold-out validation
cv: cross-validation
- autointent.custom_types.WeightType#
Represents weight calculation methods
“uniform”: Equal weight for all elements.
“distance”: Weights based on distance.
“closest”: Prioritizes closest elements.